I’m attempting to develop/train a deep learning model to pick the same lines that I think are key when extracting patterns from photos.

Jetson Nano hardware isn’t supported by folding@home, or it would be cranking away on covid proteins. But it can’t, so my Nano can work on my little art project. It’s far from achieving the goal as yet, but some of the results from this latest training run are intriguing. Unfortunately, the results from the other training examples at this point aren’t nearly as pleasing.

Day 1 in the Western Quarantine Zone

Following the governor’s order, and the ever-increasing warnings, grocery stores shelves are being picked clean for certain items like bottled water, toilet paper, and non-perishable food stuffs. Likely, this is a transient effect that will normalize as restrictions stay constant or relax as they should.

As far as the effectiveness of the order, I wonder if it is premature for my area, in Sacramento. If the quarantine is enacted too early and then lifted, the peak infection will be offset to a later date rather than smoothed out over time. Of course, the quarantine is not 100% effective, and people working in essential sectors will still be exposed, as will be people utilizing those services, so there is some possibility of a smoothing effect.

Meanwhile, I’m loving tensorflow 2. It is a dream to use compared with the old tensorflow object detection api. Tensorflow, keras, and the special processor architectures that they take advantage of open the door for whole new approaches to problem solving, which will make problems that could once only be tackled by dedicated mathematicians accessible to competent engineers (as long as they’ve access to the processing power).

My Novel Process

My novel manufacturing process is beginning to show more signs of benefit again. We had expanded it and modified it slightly and had difficulty reproducing the full benefits of the initial results, and in the process of that I’ve expanded my own knowledge of the greater system, developing an electrochemical model that I think will inform all of my decisions to come.

But now, once again, there are good and hopeful signs, and my process returns to favor in my eyes, though… there is a tinker or two I think’ll help, and if not the process, the attempts will surely help my knowledge.

A Novel Process

I invented a novel process improvement at work and the initial test results are promising, and warrant further investigation. This could end up being a big win for the company.

What makes a novel process improvement? Well I’m also working on a machine, and, although I won’t release any details, let’s say that it reduces the amount of rework required for a part of a manufacturing process. That makes the machine a process improvement, but it is not a novel process improvement, because we could pay other engineers to supply a machine that does the same thing. My machine may contain parts that are novel, but it’s not a novel process improvement.

In the case of my novel process improvement, there is no one else that can be paid to get you the same thing. Only Licap has it.

Busy, busy, busy…

I’m working mainly on a few exciting work projects these days. Mainly designing a new machine, also working with machine vendors on a new pilot line, and supporting maintenance and upgrades of existing machines, not to mention some forays into IT support.

Parts fab is at the forefront of my concerns at the moment: Do I print?, have printed?, have milled?, machine myself with or without CNC scribing? Or just all of the above? All of the above sounds right.

Cheap Aliexpress Nanos.

I got some arduino nano replicas on aliexpress a couple of months ago for an incredibly low price. Not surprisingly, there was a reason for the discount. The pin assignments as they’d be programmed in the IDE did not match the pins as they were silkscreened on the board. The analog pins were silkscreened in reverse order to their actual hardwired positions.

Thus, among the plethora of reasons that a device might end up super cheap on ali-x are manufacturing defects. This problem happenned to just be with the silkscreen, but the silkscreen is a very important time saver for circuit assembly.

Nevertheless, the board I used still helped me blast together a device to run an experiment and log data. The physical chemistry involved has been more problematic than anything else for that experiment, but a change in approach will eliminate all such problems.